Environ. Sci. Technol. 2006, 40, 3355-3361
Continuous Ice Core Melter System with Discrete Sampling for Major Ion, Trace Element, and Stable Isotope Analyses E R I C H C . O S T E R B E R G , * ,†,‡ MICHAEL J. HANDLEY,† SHARON B. SNEED,† P A U L A . M A Y E W S K I , †,‡ A N D K A R L J . K R E U T Z †,‡ Climate Change Institute, Sawyer Environmental Research Center, and Department of Earth Sciences, Bryand Global Sciences Building, University of Maine, Orono, Maine 04469
We present a novel ice/firn core melter system that uses fraction collectors to collect discrete, high-resolution (32 trace elements by inductively coupled plasma sector field mass spectrometry (ICP-SMS), and stable oxygen and hydrogen isotopes by isotope ratio mass spectrometry (IRMS). The new continuous melting with discrete sampling (CMDS) system preserves an archive of each sample, reduces the problem of incomplete particle dissolution in ICP-SMS samples, and provides more precise trace element data than previous ice melter models by using longer ICP-SMS scan times and washing the instrument between samples. CMDS detection limits are similar to or lower than those published for ice melter systems coupled directly to analytical instruments and are suitable for analyses of polar and mid-low-latitude ice cores. Analysis of total calcium and sulfur by ICP-SMS and calcium ion, sulfate, and methanesulfonate by IC from the Mt. Logan Prospector-Russell Col ice core confirms data accuracy and coregistration of the split fractions from each sample. The reproducibility of all data acquired by the CMDS system is confirmed by replicate analyses of parallel sections of the GISP2 D ice core.
Introduction Glaciochemical data from ice cores provide valuable information about past climate variability and atmospheric chemistry and their controlling forces. For example, time series of major ion (including Na+, Ca2+, Mg2+, K+, Cl-, CH3SO3-, NO3-, and SO42-) concentrations have revealed dramatic and abrupt changes in atmospheric circulation during glacial and interglacial conditions and provided insight into their forcing mechanisms (see ref 1 for a review). Stable oxygen and hydrogen isotopes in snow and ice (δ18O and δD) contribute information about past temperature, moisture source regions, and hydrology (e.g., refs 2-5). Trace elements have received more attention in recent years because of their usefulness in determining aerosol sources (e.g., refs 6-8), * Corresponding author phone: (207) 581-2112, fax: (207) 5819390; e-mail:
[email protected]. † Climate Change Institute. ‡ Department of Earth Sciences. 10.1021/es052536w CCC: $33.50 Published on Web 04/08/2006
2006 American Chemical Society
understanding biogeochemical cycles (9), and assessing the impact of anthropogenic (e.g., refs 10-12) and volcanic (e.g., refs 13 and 14) emissions on the past and present atmosphere. The conventional technique for sampling ice cores for major ion and trace element analyses includes progressively chiseling away the potentially contaminated outer portion of the core under extremely clean conditions (15). Pioneering studies utilizing this technique provided the first accurate chemical time series from snow and ice samples and established their value as paleoenvironmental proxies, but they necessarily provided low-resolution (10-20 cm/sample) and often discontinuous data sets because the method is both time-consuming and labor-intensive. The development of continuous ice core melting systems (16-24) over the past decade has reduced sample preparation time and increased sample resolution while providing continuous and coregistered data for a large suite of elements. Previous ice core melter systems have utilized in-line continuous flow analysis (CFA) techniques (16, 17) or coupled the melter to ion chromatographs (IC; 18, 19) to measure a large suite of organic and inorganic ions at ∼1 cm resolution. Recently, ice core melter technology has advanced to include continuous, online measurements of oxygen and hydrogen stable isotopes in ice core meltwater, as well as oxygen, nitrogen, and argon isotopes in air trapped within the ice core bubbles (20-22). McConnell et al. (23) coupled an ice core melter directly to an inductively coupled plasma mass spectrometer (ICPMS) and CFA system to provide continuous measurements of an extensive suite of major and trace elements. Knu ¨ sel et al. (24) expanded the number of elements measured by the McConnell-type CFA-TE (continuous flow analysis with trace elements) system by incorporating an ICP-sector field mass spectrometer (ICP-SMS) but found that 18 trace elements commonly associated with silicates (including Al, Fe, Pb, U, and lanthanides) were underestimated when compared to identical samples prepared by the conventional chiseling technique. They attribute this discrepancy to incomplete dissolution of the silicates due to limited exposure to acid (99.99% Ni) melter plate mounted on a heated aluminum base. The Ni melter plate is 3 mm thick with a series of radiating 200 µm wide slits through which meltwater flows (Figure 1). The slits produce a downward capillary force that prevents wicking of the meltwater into unmelted core, thereby allowing both firn and ice to be melted (17). The melter head splits the meltwater into two channels physically separated by a ridge (1 mm high) on the surface of the Ni plate. Meltwater from the potentially contaminated outer portion of the ice core is directed to the “outer channel” for stable isotope analysis, and meltwater from the uncontaminated, innermost portion of the ice core is directed to the circular “inner channel” for major ion and trace element analyses (Figure 1). A 1 mm high circular separator ridge on the bottom of the Ni plate nests tightly into a poly(tetrafluoroethylene) (PTFE) insert, and the PTFE insert nests tightly within the heated Al base (Figure 1). A Ni270 “cone” is sandwiched between the PTFE insert and the center of the inner channel on the Ni plate to draw the meltwater down into the PTFE insert. The Ni melter plate, Ni cone, and PTFE insert provide a clean conduit through which meltwater flows from the inner channel such that it is never in contact with the Al base. A threaded Al ring screws onto the Al base and holds the complete melter head assembly together. Thus, the Ni melter plate, Ni cone, Al base, and PTFE insert can be separated for easy cleaning between melting sessions (Figure 1c). We have constructed several different melter head assemblies to accommodate slabs as small as 2.7 cm × 2.7 cm up to whole cores as large as 10.2 cm (4 in.) in diameter. The diameter of the inner channel currently ranges from 15 mm (for 2.7 × 2.7 cm slabs) to 30 mm (for >4.2 × 4.2 cm slabs or whole core) on our various Ni plates. The melter plate is machined from Ni270 because of its favorable thermal properties, availability in pure form, reasonable cost, and because Ni sample and skimmer cones are used in the UMaine ICP-SMS, already compromising measurements of low-level Ni. Testing indicates that a melter plate coated with metal, ceramic, or fluoropolymers is problematic because the coating degrades over time, exposing the base metal underneath and potentially contaminating the sample. A solid, pure Ni melter plate can degrade without ever contaminating the sample. In addition, applying any coating within the narrow slits of the Wagenbach-style melter head (17) is technologically difficult and expensive. 3356
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 10, 2006
Ice Core Melting with Discrete Sampling. The entire UMaine CMDS system is housed in a dedicated clean room with high-efficiency particulate air (HEPA)filtered air. Nonparticulating suits and booties, facemasks and wrist-length polypropylene (PP) gloves are worn during melting and all ice preparation. Deionized water from a MilliQ-Element system (Millipore, Milford, MA; >18.2 MΩ; hereafter “DI water”) is pumped through the entire melter system between melting sessions to keep the system clean, and DI water blanks are collected prior to every melting session to confirm the cleanliness of the system. The melter head is heated to a constant temperature of 15-25 °C (depending on ice/firn density) by two 250-W cartridge heaters (Figure 1c), resulting in a 1.5-3 cm/min melt rate. The temperature of the melter head is measured by a thermocouple (Figure 1c) and maintained by an Omega Engineering (Stamford, CT) timeproportioning temperature controller. The ice/firn core or slab (hereafter referred to simply as “ice core”) is mounted upright in a 1 m long, precleaned Plexiglas container open only toward the bottom where the ice is in contact with the Ni melter plate. A PTFE-coated 250 g weight is often placed on top of the slab/core when melting firn to ensure a consistent melt rate. A high-density polyethylene (HDPE) stand holds the Plexiglas container in position over the melter head assembly. The melter head and core are housed within a dedicated upright freezer maintained at -20 °C within the clean room, while the pumps and fraction collectors described below are all located outside the freezer within the clean room (Figure 2). Meltwater from the uncontaminated inner channel is pumped from the melter head through PTFE tubing (1.6 mm i.d., 3.2 mm o.d.) and platinum-cured silicone/PTFE pump tubing (STA-PURE, Gore and Associates, Elkton, MD; 2.8 mm i.d.) by two Gilson (Middleton, WI) peristaltic pumps. A third peristaltic pump controls the flow of meltwater from the outer channel. Pump rates are dictated by the ice melt rate and ice core dimensions and are adjusted so that there is always net flow from the inner toward the outer part of the core to prevent contamination of the inner sample. PTFE tubes are flanged and connected to the Al melter base (for the outer channel) and PTFE insert (for the inner channel) with a threaded LDPE fitting (Figure 1c). DI water blank sample volumes are checked prior to each melting session to monitor any pump tube degradation, although the STAPURE pump tubes have exceptional longevity (several months) compared to silicone or Tygon peristaltic pump tubes (days).
FIGURE 2. Schematic diagram of the continuous melter with discrete sampling (CMDS) system. Three fraction collectors collect discrete samples from a continuous meltwater stream for IC, ICP-SMS, and IRMS analyses. The entire CMDS system is housed in a dedicated clean room with HEPA-filtered air. Meltwater from the inner channel is split between two Gilson fraction collectors: one that collects ICP-SMS samples in acid-precleaned, PP vials under a class-100 HEPA clean bench, and a second fraction collector that collects IC samples in PP vials precleaned with DI water. Meltwater from the potentially contaminated outer channel is collected by a third Gilson fraction collector in 20-50 mL PP or high-density polyethylene (HDPE) vials for stable isotope analysis. The “master” fraction collector measures the IC sample volume by counting drops, and triggers the ICP-SMS and stable isotope fraction collectors to simultaneously advance to the next sample when the predetermined volume is collected. The volume collected for each fraction depends on the melter head and ice core dimensions, number of analytes, and the desired sample resolution. A minimum of 4 mL is generally required for IC analyses, and 2 mL is desirable for ICP-SMS analysis. These requirements result in a maximum depth resolution of ∼1 cm ice/sample (∼2 cm/firn sample) when the Ni melter plate with a 30 mm inner channel diameter is used, assuming that both IC and ICP-SMS data are desired. Typically we collect larger sample volumes to preserve an archive for subsequent analyses. Higher sample resolution is possible with different melter head dimensions (i.e., larger inner channel diameter) and/or restricting inner samples to either IC or ICP-SMS analyses (see Efficiency of Ice Decontamination below); however, natural inhomogeneity in the ice may limit the usefulness of 0.99. IRMS Analyses. Stable isotope analyses (δ18O and δD) of meltwater samples are performed via gas source mass spectrometry. Oxygen isotope ratios are measured via standard CO2 equilibration techniques with a Micromass Multiprep device coupled to a stable isotope ratio analysis (SIRA) mass spectrometer ((0.05‰ precision based on longterm standard analysis statistics). Hydrogen isotope ratios are measured via Cr reduction with a Eurovector elemental analyzer coupled to a Micromass Isoprime mass spectrometer ((0.5‰ precision based on long-term standard analysis statistics; 26). Data are reported in delta (δ) notation relative to standard mean ocean water (SMOW).
Results and Discussion Accuracy, Precision, Blanks, and Detection Limits. Although no standard reference material currently exists for trace elements in polar ice, analyses of the reference materials (Environment Canada) SLRS-4 by ICP-SMS and ION-92 by IC produce concentrations within the certified range (Table 1), confirming accuracy. Instrument detection limits for the IC and ICP-SMS, equivalent to 3 times the standard deviation (3σ) of 10 DI water samples, are listed in Table 2 along with VOL. 40, NO. 10, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3357
TABLE 1. Comparison of Standard Reference Material (SLRS-4, ION-92) Certified Values (Environment Canada) with Values Determined from the UMaine ICP-SMS and IC element Na+ b Ca2+ b Mg2+ b K+ b Cl- b SO42- b Al As Ba Ca Cd Co Cr Cu Fe Mn Pb Sr U V b
UMaine analyzeda (µg/L)
certified (µg/L)
20.2 ( 0.4 44.4 ( 1.8 10.7 ( 0.5 1.1 ( 0.2 106.0 ( 3.4 37.8 ( 2.5 53 ( 2 0.70 ( 0.02 12.0 ( 0.5 6.4 ( 0.2 0.012 ( 0.001 0.030 ( 0.003 0.35 ( 0.01 1.68 ( 0.07 106 ( 4 3.41 ( 0.10 0.081 ( 0.001 27.6 ( 0.9 0.050 ( 0.004 0.30 ( 0.01
19.3 ( 2.4 42.8 ( 4.0 9.6 ( 1.1 0.9 ( 0.2 105.5 ( 12.3 37.0 ( 4.9 54 ( 4 0.68 ( 0.06 12.2 ( 0.6 6.2 ( 0.2 0.012 ( 0.002 0.033 ( 0.006 0.33 ( 0.02 1.81 ( 0.08 103 ( 5 3.37 ( 0.18 0.086 ( 0.007 26.3 ( 3.2 0.050 ( 0.003 0.32 ( 0.03
FIGURE 3. Concentration cross-sections of Mt. Logan PR Col firn created by melting the firn sideways through the CMDS system at 3 mm/sample resolution. Only the outer 3-6 mm is contaminated.
a Values represent mean of 10 analyses. Errors correspond to 1σ. IC data (ION-92).
average DI blank concentrations. These detection limits and DI blank values are similar to or lower than published values on identical or comparable instruments (e.g., refs 24, 25, and 27; Table S2, Supporting Information). Continuous melter method blanks and detection limits, calculated from 10 DI water blanks passed through the entire CMDS system, are listed in Table 2 and compared to Greenland ice concentrations (10, 28-31; unpublished UMaine data) from samples spanning the Holocene and glacial periods. CMDS detection limits are similar to or lower than published CFA-TE detection limits (24; see Table S2, Supporting Information), and are almost always lower than minimum Greenland ice values and an order of magnitude below average Greenland ice values (Table 2). The δD ratio of a stable water isotope reference material (homogenized Antarctic surface snow) that was frozen and melted through the CMDS system (-202.4 ( 1.2‰, mean and 1σ of 10 samples) matches its δD ratio before melting (-201.8 ( 0.5‰), confirming the accuracy of stable isotope measurements. We have discovered that Ni from the melter plate elutes close to Ca2+ on the CS-12A IC column. The two peaks are easily distinguishable when both are present (in core meltwater samples), but the Ni peak can be mistakenly interpreted as Ca2+ in method blanks when the true Ca2+ concentration is below the detection limit. This can lead to erroneously high concentrations of Ca2+ in melter blank tests when analyzed solely by IC. Efficiency of Ice Decontamination. Figure 3 displays examples of concentration cross-sections generated by melting small pieces of firn sideways through the CMDS system at high resolution, analogous to cross-sections generated by physically removing successive veneers of ice and analyzing each veneer (15, 24). The firn is from the 2001 Prospector-Russell (PR) Col core collected on the summit plateau (5300 m elevation) of Mt. Logan, Yukon, Canada. Each data point in the cross-sections represents 3 mm of firn, which was achieved by sampling only for ICP-SMS analyses. In all cases, contamination is limited to the outermost 3-6 mm of the slab, and stable low-concentration plateaus are observed. Consequently, a 6 mm or larger buffer between the slab/core edge and the inner meltwater channel is always maintained when melting. 3358
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 10, 2006
FIGURE 4. CMDS calcium and sulfur profiles from the Mt. Logan PR Col ice core (∼1893-1906 AD) showing exact coregistration of IC and ICP-SMS samples. Sample resolution is ∼1.9 cm/sample (14-22 samples/year), and the data have not been smoothed. McConnell et al. (23) and Knu ¨ sel et al. (24) cite meltwater dispersion within the melter head, tubing, and debubblers as limiting sample resolution of the CFA-TE system to ∼1 cm/sample. The ice decontamination data presented here, however, suggest that mixing does not significantly distort data at a resolution of 3 mm/sample. The PTFE tubing contains ∼1 sample worth of meltwater (with the standard sample sizes of 4 and 2 mL for IC and ICP-SMS, respectively) during melting, minimizing any effects of mixing in the system. Sample Coregistration. We currently measure Ca2+, SO42-, and CH3SO3- on the IC and elemental Ca and S on the ICPSMS, providing two time series with which to confirm sample coregistration. Elemental sulfur concentrations from the ICPSMS can be directly compared to IC data by dividing SO42by 3 and CH3SO3- by 2.97 and summing their concentrations. Figure 4 shows that Ca and S concentrations from the Mt. Logan PR Col core measured on the two instruments are very highly correlated (R2 ) 0.88 for Ca, R2 ) 0.97 for S) confirming sample coregistration. This comparison also validates the concentration values produced by the two different instruments. Reproducibility of CMDS Data. Parallel 2 m long slabs of ice from the GISP2 D core (meters 2023-2025; ∼25 800 BP; 31, 32) were melted to demonstrate the reproducibility of chemical measurements made with the CMDS system. IC, ICP-SMS, and stable isotope data are displayed in Figure 5 and compared to lower resolution (20 cm/sample) IC data processed with traditional scraping methods and/or an early version of a continuous ice melter (31). Outliers due to noise spikes in the mass spectra caused by small particles (24) have been removed from the ICP-SMS data, but the data
TABLE 2. Average Instrument and Procedural (CMDS) Blanks, and Instrument and Procedural Detection Limits (DL) Compared to Greenland Ice Concentrations (ng/L) analytea
DI water blankb
inst.DLc
CMDS blankd
CMDS DLe
Greenland concn min-max, (mean)f
Na+ (IC) Ca2+ (IC) Mg2+ (IC) K+ (IC) CH3SO3Cl- (IC) NO3- (IC) SO42- (IC) 27Al (MR) 138Ba (LR) 209Bi (LR) 44Ca (MR) 111Cd (LR) 59Co (MR) 52Cr (MR) 133Cs (LR) 63Cu (MR) 56Fe (MR) 55Mn (MR) 208Pb (LR) 32S (MR) 88Sr (LR) 47Ti (MR) 205Tl (LR) 238U (LR) 51V (MR) 139La (LR) 140Ce (LR) 141Pr (LR) 146Nd (LR) 147Sm (LR) 153Eu (LR) 157Gd (LR) 159Tb (LR) 163Dy (LR) 165Ho (LR) 166Er (LR) 169Tm (LR) 172Yb (LR) 175Lu (LR)
